Multi-nesting spatial variability and source analysis of statistical technology were applied to the typical plots which consist of brunisolic soil. By means of the sample testing results on five elements of heavy metals: Pb, Cd, Hg, Cu, Cr, with As element, which could reflect the levels of the soil pollution, aforementioned geo-statistical analysis process was done. Optimal model was selected by kriging interpolation method. And then, the reasonable sampling spacing was well determined. Combined with semivariance and fractal dimension analysis, the reasonable sampling scale was obtained. The method combined PCA with semivariance analysis together to trace to the source was adopted forward to guide agricultural management measures. The results indicated that the origin of farmland soil heavy metal composition was composed of three main source formations, which were soil parent material, precipitation and irrigation, and pesticide application. Medium scale jibed with semivariance model fitting. The value of range varied from 319m to 752m. The spherical model was thought more suitable to describe space distribution of soil heavy metal. 